Room: Exhibit Hall | Forum 9
Purpose: HPV-associated oropharyngeal squamous cell carcinoma (OPSCC) can be treated with definitive radiotherapy (RT), combined chemoradiotherapy (CRT), or transoral robotic surgery (TORS) which provides comparable outcomes (OS rate of 80%) with potentially lower toxicity profile. However, approximately 33% of the TORS patients ultimately require adjuvant CRT (tri-modality therapy with the highest toxicity profile) because of finding such as extracapsular extension and positive margins on the final pathology. This study investigated CT textures predictive ability for least favorable candidates for TORS, who would ultimately receive tri-modality therapy.
Methods: 92 consecutive TORS patients treated in our institution were selected. Pre-TORS diagnostic CT, contoured by the attending radiation oncologist, indicated 61 cases of enlarged lymph nodes. In addition, the nodes were expanded by 0.3 cm and a 0.5 cm shell was also outlined around the node periphery. 126 geometric-, first-, second-, and third-order textures were extracted from each contoured structure per patient. The patient cohort was dichotomized to: 1) TORS alone and TORS plus RT, and 2) TORS plus CRT. The textures were associated with the primary endpoint - group 1 vs group 2. A univariate logistic regression model was used to estimate statistical significance, and Cohenâ€™s d was calculated for effect size assessment.
Results: Only textures for the originally contoured nodes demonstrated significant associations with treatment approach. 14 features for the nodes were significant (p-values = 0.015-0.047). 4 features had small effect (d < 0.2), 8 features had small-to-medium effect (0.2 < d < 0.5), and 2 features had medium-to-large effect (0.5 < d < 0.8).
Conclusion: CT textures on pre-surgery imaging may help in the treatment selection of HPV-associated OPSCC patients, specifically those considering TORS who would ultimately require tri-modality therapy. Future directions for research include harmonizing the CT data, inclusion of clinical variables, and validation through multivariate analyses.